Fold hourly 0445 HIGH

[24601]

Hourly 0445 HIGH fold after merged #63 (0348 / notes.md §139). Fresh PR off latest main. Does not bump 0.5.0. Do not reopen or amend PR #23–#63.

What this fold does

Catalogs the 0445 Boise watch (49 novel+revisit HIGH) into notes.md §140 / composition items 649–664 / findings batch #122. Uniqueness lock UNIQ_0445 is one consecutive substring across the overlay set. Third-party benches stay theirs.

Primary cards

• lucasmartins-ai/lcc: real Laya keeps essentially every block (0.0% / −0.5% theirs). Mechanical −70.0% / Jev −52.1% theirs. Mock “Laya = Jev, −22.6% on XL” withdrawn. Token reduction alone is not cost reduction. N=18 pilot not Harbor.
• David-Lolly/Jev-Compatible: Chat 1282.3 ms vs gateway 232.1 ms (~1/5.5 theirs). Softmax over candidate logprobs. Qwen3.8-27B ≠ Archer. wire-compat ≠ logit-equiv.
• hwfengcs/any2jev: independent, not affiliated. acc 0.796 ECE 0.027 theirs. 42 ms vs JSON 778 ms theirs. Qwen3-0.6B ≠ Archer.
• TianyuCodings/NanoJev densify §115: README now links JevHarness (HEAD 618cea6d → 76fdfc9e). densify §115 not a sibling first sighting. SHA move is not a replica.
• jjd-lab/jev-synthetic-survey: How you ask mattered more. Noul TVD 0.1530 vs GPT 0.1789 theirs. ECE 0.1472 missed the 0.05 bar.
• CankatSarac/jev-arcade: snake 70/80 theirs. tetris 167 vs heuristic 2333 theirs. Calibration is not yet measured. three seeds not Harbor.

Remainder densify (this revision)

Named first cards for the 21 leftover HIGH repos that had README substance, instead of a dump-wall remainder. Class locks: not a digital twin (traffic sim, Mean wait 24.15 s theirs); Score fan-out ≠ chess engine (xiangqi Opening 44 moves $0.000322 6.6 s theirs); Jev cannot waive a failing check (KorWF-Pi). Namesakes: umgbhalla/jevx ≠ hawkyre/jevx; ryanzen9/XFlow ≠ hawkyre/jevx; Tsagaanbayr1/jev-tetris ≠ planstack-ai/jev-tetris-benchmark; wizicer/jev_info_site ≠ JingHao-Leon/awesome-jev-apps. UNIQ_0445 stays frozen (4127 chars).

Also this hour

• First cards: AXERA-TECH/Laya (AX650 NPU theirs), openjev-MLX-4bit (~15 GB 4-bit), GeekyAbs/laya ≠ convaiinnovations/laya, laya-needle, laya2typesafeapi, iamdgarcia/openJev, homebrew tap for laya-mac-serve §139, judgements, awesome-jev-apps, jev-chat, jev-snake-game, jevgo, sszxt/rlcd (ECE 0.490→0.423 theirs).
• Census watch (not re-derived): Awesomejev 691→802 (+111) / 38194→52151 stars. tracker likes 81, lastModified UNCHANGED.
• skip-thin empty/thin README: jev-spire-brain, jev-robotic-arm-benchmark, jev-study, jev-laya-statement, TDM-demo. danish-dynaword-laya gated HTTP 401.

Gates

• uniqueness_gate.py now checks 0843…0348 + 0445. Also requires the 21 remainder IDs in notes.md §140.
• evaluate_decisions.py --self-test adds lcc keep-all / softmax gateway / any2jev ECE / How you ask / arcade calibration / not a digital twin / Score fan-out ≠ chess engine / Jev cannot waive a failing check.
• revisit_fingerprints.py --self-test updates NanoJev SHA and first-card notes_section 140, including remainder SHAs.
• Archive: research/archive/hourly/2026-09-21T11/ (invented_signal: false).

  

[24601]

ADV_PASS. Adversarial review + testing of #64 (HEAD cc4ef425b66e65670089caffb17c56fd71908f33).

Honesty checked: lcc keep-all theirs (token≠cost; N=18 ≠ Harbor); Jev-Compatible Chat 1282.3 ms vs gateway 232.1 ms theirs (wire≠logit); any2jev acc/ECE theirs (Qwen3-0.6B ≠ Archer); NanoJev densify §115 HEAD 618cea6d→76fdfc9e (JevHarness, fingerprints notes_section still 115); remainder densify in notes §140 (traffic sim ≠ digital twin; xiangqi Score≠chess engine; KorWF-Pi Jev cannot waive a failing check; umgbhalla/jevx ≠ hawkyre/jevx; wizicer/jev_info_site ≠ JingHao-Leon/awesome-jev-apps). UNIQ_0445 frozen at 4127 chars by design; remainder IDs gated in notes §140 + SKILL protocol + evaluator. invented_signal false. No version bump. Do not reopen #23–#63.

Tests (python3, all exit 0):

• .agents/skills/augustus/scripts/uniqueness_gate.py → uniqueness-gate ok (0445 chars=4127, 21 overlays)
• .agents/skills/augustus/scripts/evaluate_decisions.py --self-test → self-test ok
• research/revisit_fingerprints.py --self-test → revisit-fingerprints self-test ok

IDs unique vs main (777546f / merged #63): §140 / items 649–664 / batch #122 (main owns §139 / 633–648 / #121). Occupancy steal FAIL-as-expected for §140, 649, 664, #122; restored. SKILL 0.5.0 frozen. Archer still promised_not_landed. CHANGELOG Unreleased has no uniqueness dump wall. CI build success. READY_TO_MERGE. Do not merge from this comment.